Context-Aware Knowledge Extraction from Legal Documents Through Zero-Shot Classification

The extraction of conceptual and terminological knowledge from legal documents is a crucial task in the legal domain. In this paper we propose ASKE (Automated System for Knowledge Extraction), a system for the extraction of knowledge that exploits contextual embedding and zero-shot learning techniques in order to retrieve relevant conceptual and terminological knowledge from legal documents. Moreover, in the paper we discuss some preliminary experimental results on a real dataset consisting of a corpus of Illinois State Courts’ decisions taken from the Caselaw Access Project (CAP).